Defining Difficult-to-Treat Rheumatoid Arthritis in Routine Care: Comparative Performance of Published Criteria and Description of Early and Late Difficult-to-Treat Rheumatoid Arthritis Phenotypes.

OBJECTIVE: Despite advances in treat-to-target strategies, a subset of patients with rheumatoid arthritis (RA) remains refractory to multiple therapies and is classified as having difficult-to-treat RA (D2T-RA). Although EULAR has proposed classification criteria for D2T-RA, these patients remain variably identified in real-world practice because of heterogeneous applications. We aimed to compare the preformance of published applications of the EULAR D2T-RA criteria in a real-world RA cohort and to determine weather the timing of D2T-RA development could distinguish early from late D2T-RA phenotypes.

METHODS: Using data from the Ontario Best Practices Research Initiative, a longitudinal real-world RA registry, we evaluated four literature-derived adaptations of the EULAR D2T-RA definition among patients initiating their first advanced therapy. These definitions differed in disease activity requirements, follow-up duration, and treatment count thresholds. We assessed prevalence, agreement between definitions, and explored different temporal anchors for classifying early versus late D2T-RA.

RESULTS: Among 1,121 patients with RA (mean age 56.0 years; 80.7% female; mean disease duration of 8.4 years), 202 (18%) fulfilled at least one of four D2T-RA definitions. Prevalence varied substantially across definitions, ranging from 5.5% (definition 4) to 11.7% (definition 2). Definitions 1 and 2, which incorporated multiple disease activity measures, identified the largest proportion of patients, whereas definition 4, which is defined solely by initiation of a third advanced therapy, was most restrictive. Overlap between definitions was modest, with only 44% of patients meeting all definitions 1 through 3, whereas definition 4 identified a distinct group with no overlap with the other definitions, suggesting that these classification approaches capture different patient subsets and are not interchangeable. In temporal analyses, anchoring the D2T-RA onset to the initiation of the first advanced therapy, a two-year threshold-selected for its clinical relevance and balanced group sizes-classified approximately 40% of patients as early and approximately 60% as late D2T-RA.

CONCLUSION: The applications of D2T-RA criteria are not interchangeable and vary widely in patient capture. A two-year threshold anchored to the first initiation of advanced therapy offered a clinically balanced early-late D2T-RA framework. Harmonized, multidimensional definitions incorporating disease activity and temporal stratification are needed to improve phenotyping, prognostication, and cross-study comparability.